Abstract:

Abstract Lung cancer, a widespread and potentially disastrous type of cancer, requires urgent action to prevent catastrophic consequences caused by delayed medical care. Currently, Computed Tomography (CT) scans are used to assist doctors in detecting lung nodules at an early stage. However, the accuracy of lung nodules diagnosis heavily relies on the expertise of physicians, which can lead to potential oversight of specific patients and subsequent difficulties. As a result, deep learning has emerged as a highly considered and effective method in various medical imaging fields, including lung cancer and nodules detection. In this paper, we proposed a Custom-VGG16 model to analyze CT images of lungs and accurately classify malignant lung nodules. The Custom-VGG16 model was tested using the LIDC-IDRI database to evaluate its effectiveness. The experimental results highlight the remarkable performance of the Custom-VGG16 network, achieving an accuracy rate of 95%. Furthermore, the results indicated that the Custom-VGG16 network outperforms both the VGG16 and CNN models in detecting lung nodules.

Keywords: